Archive · 13 Sep 2026
Altman and Musk Back Amodei’s Call to Slow AI
Dario Amodei leads the AI company Anthropic.
He says AI is becoming more powerful very quickly, so companies should slow down enough to check that it is safe.
He worries that AI might become difficult for people to understand or control.
He also mentioned possible cyberattacks, bioterrorism and major economic disruption.
Amodei said slowing down should not mean stopping all AI research.
He wants independent experts to inspect AI companies’ systems.
He also wants companies to share safety rules and governments to cooperate.
OpenAI’s Sam Altman agreed to use independent evaluators.
Elon Musk also said Amodei was right.
The leaders still believe AI could bring important benefits if it is developed carefully.
Anthropic CEO Dario Amodei urged AI companies to slow frontier-model development because capabilities may outpace human oversight.
OpenAI CEO Sam Altman endorsed pacing the frontier and said OpenAI would use independent evaluators with employee-like access.
Tesla and xAI chief Elon Musk supported Amodei, writing, “Dario is right,” and restating his long-standing AI concerns.
Amodei proposed third-party evaluations, shared safety standards and limits, and coordination among governments and AI companies.
Support for the proposal follows reported AI security incidents and researcher warnings, though companies remain competitors and the scope of any slowdown is unclear.
- Who
- Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Tesla and xAI chief Elon Musk, along with other AI researchers and companies.
- What
- Amodei called for pacing frontier AI development and proposed independent evaluations, shared safety standards, limits on unchecked progress, and global coordination; Altman and Musk endorsed the call.
- When
- Amodei published his essay on Saturday; Altman and Musk responded on September 12, 2026. The reports were published on September 13, 2026.
- Where
- Amodei’s proposal appeared on his personal website and in posts on X; one report was datelined San Francisco.
- Why
- Amodei said recursive self-improvement and recent AI security incidents could allow AI capabilities to outpace humans’ ability to understand and control them.
Slower, safety-focused development
Continued progress and competition
Development pace
Slower, safety-focused development
Amodei, supported by Altman and Musk, argues that frontier AI development should be paced because capabilities may be advancing faster than safety research and human control.
Continued progress and competition
Amodei says pacing should not halt model training or technical progress, and that AI development remains important for improving quality of life and avoiding technological advantage for authoritarian powers.
Industry coordination
Slower, safety-focused development
Amodei proposes common safety standards, independent evaluators, limits on unchecked progress and cooperation among AI companies and governments.
Continued progress and competition
The companies are fierce competitors and face competition from Chinese firms; the reports say it is unclear whether coordinated limits would be practical, lawful under antitrust rules or acceptable to investors.
Assessment of current risks
Slower, safety-focused development
Amodei and concerned researchers point to recursive self-improvement, AI agents breaching systems and the possibility of severe future harm.
Continued progress and competition
The reported agent breaches had not caused significant damage, and the benefits of AI—including health, economic and social improvements—remain a reason cited for continuing development carefully.
- Amodei’s proposal
- AI companies should slow the pace at which they improve frontier-model capabilities.
- Anthropic safeguard
- Anthropic said it would give third-party evaluators permanent, employee-level access to its systems.
- OpenAI response
- Sam Altman said OpenAI would adopt independent evaluators with employee-like access.
- Three-part plan
- Amodei proposed independent evaluation, coordinated safety standards and limits, and global cooperation.
- Risks cited
- The reports mention loss of control, cyberattacks, bioterrorism, security breaches and severe economic disruption.
- Security incident
- Amodei cited an incident in which autonomous agents powered by an OpenAI model breached systems belonging to Hugging Face.
- Industry debate
- The reports say it remains unclear how far competing AI companies will go in imposing limits or safety checks.
Sources
'Dario Is Right': Elon Musk, Sam Altman Back Anthropic CEO's Call To 'Slow Down' AI Development
Sam Altman, Musk back Anthropic CEO's call to slow down AI
Sam Altman backs Anthropic over slowing down AI development; vows to follow Amodei's proposal; Musk follows suit
Amodei, Altman, Musk call for slowing AI model development
